SA + PostFix + Virtual final question for a while

["Gerald V. Livingston II" <ge...@sysmatrix.net>]

I don't expect I'll get a reply quick enough to help. It's been strongly
suggested that I switch our mail server over to ANYTHING other than what we
have before the weekend is out.

I've been bouncing around the web and digging through documentation trying to
figure out how to do SA + virus scanning in a virtual domain SQL environment
with per-user prefs working.

I can get all the parts working individually, but not at the same time.

It looks like I'll have to go with an all Amavis install and dump the
per-user prefs except the few that Amavis incorporates in conjuction with
using Mail::SpamAssassin. I really wanted the user prefs available through
spamc/spamd/SQL to be available as those would help users cull more spam
prior to delivery into their mailboxes. Things like what charsets to allow etc.

Here's the problem. I need a way to actually DO SOMETHING with the mail in a
virtual setup after it is processed by spamc/spamd. It has to be something
that will honor the users' SQL prefs that they set up through some web
interface and that *I* don't have to manually set up and maintain for each
user.

In a non-virtual setup I'd probably use procmail. Since the usernames will
not correspond with system accounts that's not really viable.

I need something that will allow me to set a system wide cutoff of say 30 or
40 (or 20 if that looks good later) and will send everything that exceeds
that score to /dev/nul and will also send to /dev/nul and messages exceeding
the individual users' selected cutoff score without ever having delivered it
to their inbox. 99% of our users will be POPing their mail off of the server
and they don't want to download blatant spam to be processed by their MUA
filters (actually, 98% don't have a clue as to how to set up a filter in
their MUA).

The only IMAP access they will have to their local mail store is via
Squirrelmail or IMP. Port 143 will be firewalled otherwise (because 97% of
our users would leave loading IMAP connections open on the server the entire
time they were connected if we made IMAP access available --- and told them
how to set up IMAP in their MUA of course). We won't be using Bayes for this
reason. Too difficult to get clean messages back for training. We most likely
won't use ASS (it fits, keep it) at first either.

So, I want to:

1. recieve via Postfix port 25

2. pass to Amavis port 10024 for virus scan

3. pass non-viral messages back to postfix port 10025

4. pass to *something* on 10026 that will invoke spamc and DELETE unwanted
high scoring spam (spamd should add header tags to EVERYTHING -- whatever the
smtp wrapper program is should /dev/nul the unwanted ones)

5. pass non-spam back to postfix port 10027 for final delivery to the virtual
maildir.

I have to have *something* in place by Monday morning and I still have to
extract 6000+ username/password pairs from an Access database and import them
to whatever the final form of the MySQL database tree is here -- with
intelligent defaults for the other fields.

So, do I just go with Amavis and it's limited subset of per-user prefs or is
there another choice I cam implement on such short notice?

Thanks all. I'm removing all the various packages I installed for testing as
I write this. Sometime tomorrow I need to do a clean install/upgrade to
whatever will be the final product.

Gerald